It looks as if there has been a gradual tendency for Programming messages to get shorter over during the period for which I could extract monthly archive data, from October 2005 to present. A plot of my results, which is deseasonalized as be for, using the (CMA) centered moving average can be viewed at the following link. [Actually it is more common for CMA's to be computed using the mean rather that the median, but I am partial to the median.] The contents of the temp/ files below have been posted earlier so I have not shown them here.
http://www.pixentral.com/show.php?picture=130j3mHm7m0MfkyCzBLkowUhjU66 My analysis went as follows. load 'plot' load 'stats' load '/Users/brian/j602-user/temp/13.ijs' $programming 2682 $p =: programming 2682 $size =: |.". @((_5}.(>:@i:&'t')}.]));._2 p 34 load '/Users/brian/j602-user/temp/16.ijs' $data =: readdata yearmonth 5 10 34 34 plot size%~data plot }:size%~data 'xtic 12' plot -:((12 median \])@}:+(12 median \])@}.)}:size%~data ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
